Make Enough Money in 20 Days to Cover a Whole Year! Moonlight Kimi K2.5 Soars: Overseas Revenue Surges, Valuation Exceeds Billions and Securely Holds the Position of the Ten-Horned Beast

The "unicorn" domestic large model's profit-making speed has once again refreshed industry perceptions. On February 24, 2026, according to OSCHINA, the latest release of Moonshot AI, Kimi K2.5, has shown astonishing profitability: within just 20 days of its launch, its total revenue has officially exceeded the total for all of 2025.

Core Breakthrough: Overseas Paid Users as "Patrons"

The performance miracle of Kimi K2.5 is not accidental; it is the initial victory of a global strategy:

Payment Surge: The core driving force comes from a surge in global paid users, especially a significant increase in API call volume.
Flowers Bloom Outside the Wall: Data shows that overseas users have made a particularly outstanding contribution to Kimi K2.5's revenue. On the OpenRouter platform, the call volume of Kimi K2.5 continues to rank among the top in the industry, proving its competitiveness in the international developer ecosystem.

Rapid Rise in Value: The Fastest "Decacorn"

In the capital market, Kimi is also leading the way:

Valuation Leap: After completing over 1.2 billion USD in consecutive financing, Kimi's latest valuation has soared to between 10-12 billion USD.
Record Breaker: This makes it the fastest company in China to reach the "Decacorn" (a company valued at over 10 billion USD) stage since its founding.

Leader's Statement: Holding a Large Amount of Funds, No Urgency to Go Public

Facing external attention, founder Yang Zhilin expressed great calmness in an internal letter:

Abundant Funds: The company's current funding scale has already exceeded the fundraising amount of most IPOs or private placements.
Strategic Patience: Due to the extremely abundant funds in the primary market, Kimi is currently focusing on technological iteration, and there are no plans to go public in the short term.

Technical Profile: Kimi K2

Model Architecture: Mixture of Experts (MoE) architecture.
Parameter Scale: A total of 1 trillion parameters, with 32 billion activated parameters.
Performance Optimization: Based on the MuonClip optimizer, it has achieved a qualitative leap in long-text processing and reasoning efficiency.

Honor Enters Embodied Intelligence: Its First Humanoid Robot Will Debut at MWC2026

Honor will unveil its first humanoid robot at MWC2026, targeting consumer markets such as home services and intelligent companionship. This move marks a key step in its transformation toward an AI terminal ecosystem. To support this strategy, the company established a new industry incubation department in 2025, which includes five laboratories, and plans to invest $10 billion over the next five years.

Chinese Large Models Dominate OpenRouter: Market Share Exceeds 60% During Spring Festival, Programming and Intelligent Agents Become Key Battlefields

During the 2026 Spring Festival, Chinese AI large models showed outstanding performance in the global developer ecosystem. According to data statistics from AIbase on the OpenRouter platform, the total token consumption of Chinese models in the top ten accounted for as high as 61%, taking the top three positions, indicating strong market influence. The top three models are MiniMax M2.5, Kimi K2.5, and Zhipu GLM-5, with MiniMax M2.5 leading the ranking with 24.5 trillion token calls.

Anthropic Launches Employee Stock Sale Program, Total Value May Reach 6 Billion Dollars

Anthropic has launched a large-scale employee stock sale program, with expected proceeds of up to 6 billion dollars, targeting current and former employees who have worked for at least one year. This transfer is based on the company's latest valuation of 380 billion dollars, which represents a further significant increase from the 3 billion dollars valuation after the company completed its financing at the beginning of this month.

Performance Approaching Flagship with Increased Capacity Without Price Hike: Anthropic Launches Claude Sonnet 4.6 Supporting One-Million-Token Long Text

Anthropic released Claude Sonnet 4.6, achieving significant improvements in core capabilities while maintaining a mid-range price, even surpassing the previous flagship model. The key highlight is introducing a context window of up to 1 million tokens (in Beta phase), enabling it to process an entire complex codebase or lengthy document in one go.